Measurement error in computed tomography pelvimetry.

Computed tomography pelvimetry is still used in clinical practice. We wished to quantify observer error in order to assess the level of confidence with which pelvic measurements can be described as adequate or inadequate. Anteroposterior inlet, anteroposterior outlet, transverse inlet and interspinous distances were measured from 11 CT pelvimetry examinations by five observers at one institution. Three CT pelvimetries were measured by five observers at a second institution. Intraobserver and interobserver variation was assessed using analysis of variance. Reliability of measurements was assessed using intraclass correlation coefficient. Combined error was calculated to determine 95% confidence limits for published minimum recommended pelvic measurements. The standard error of measurement, combining all sources, for measurement of the bony dimensions of the pelvis were: for anteroposterior inlet, 2.0 mm; anteroposterior outlet, 6.9 mm; transverse inlet, 1.3 mm; and interspinous distance, 2.1 mm. The 95% confidence interval around the recommended anteroposterior outlet of 100 mm was 88.5-111.3 mm. Observer variation in measurement of anteroposterior outlet is so large as to make the measurement of doubtful clinical utility.